Correlation Between MRI Findings and Clinical Impairment In Patients With Shoulder Adhesive Capsulitis: A Prospective Study

MRI diagnostic criteria of shoulder adhesive capsulitis (AC) are widely used, but there is little information available on the association between MRI ndings and clinical impairment. The aim of our study was to determine the correlation of MRI ndings with the Constant-Murlay Score (CMS), pain duration, and symptoms at the one-year follow-up in AC patients. Methods: MRI of 132 patients with a clinical diagnosis of shoulder AC were prospectively studied. A radiologist examined all patients and completed the CMS just prior to MRI. Pain duration was assessed. The signal intensity and the maximal thickness of the inferior glenohumeral (IGHL) and coracohumeral (CHL) ligaments were measured by two radiologists. Medical record analysis was performed in a sub-group of 49 patients to assess correlation approximately one year after the MRI examination. There was a difference in mean pain duration score (3.8 1.2 versus and versus respectively and and in (15.7 versus ± and and a high compared low <


Introduction
Adhesive capsulitis (AC) of the shoulder is a common condition with an incidence in the general population varying considerably from 2 to 5.3% for primary and from 4.3 to 38% for secondary AC (e.g., AC preceded by a clinical or surgical event) 1 . Although spontaneous resolution is the rule, years can ensue (mean 18-30 months) before joint mobility returns to normal 2 . Various treatment options exist for AC (e.g., oral anti-in ammatory drugs, intraarticular corticoid injection, physiotherapy, percutaneous capsular distention, surgical release, etc.) depending on the level of clinical impairment, and on an accurate diagnosis. Thus, disease staging and identi cation of in ammatory changes could have an impact on patient management 3 .
The diagnosis of AC is classically based on clinical presentation, medical history, and physical examination. Diagnosing this condition, however, can be challenging as AC may occur in various clinical scenarios and has multiple potential differential diagnoses (2,4). Imaging plays an ever-growing role in the evaluation of patients with suspected AC ruling out pathologic conditions that can clinically mimic AC 4 , and in diagnostic con rmation when clinical ndings are equivocal [5][6][7][8][9][10][11][12][13][14] . AC suggestive MRI ndings are well recognized and primary involving the inferior glenohumeral ligament (IGHL) (hypersignal and thickening), and rotator interval (RI) 7,9,10,12,15−17 . Patients with AC typically complain of a gradual and progressive onset of pain, sleep-disturbing night pain, and painful active and passive limitation at various degrees of ranges of motion (ROM), for at least 1 month 18 . The Constant-Murley Score (CMS) is often used to evaluate the impact of AC in shoulder function, with potential implications in patient management 19 . Although the correlation of MRI ndings with clinical staging was demonstrated in 2008 by Sofka et al. 20 , there is little information available on the association between MRI ndings and clinical impairment, which could be important for therapeutic decision making 15,21−26 . We hypothesize that IGHL and coracohumeral ligament (CHL) signal and thicknesses are associated with the degree of shoulder function impairment and AC progression over time. The aim of our study was to evaluate the correlation between MRI ndings in AC patients, CMS, and symptoms at the one-year follow-up.

STUDY GROUP
The institutional review board of the CHRU of Nancy, FRANCE, approved this study, and all patients gave written informed consent. All experiments were performed in accordance with relevant guidelines and regulations. From October 10, 2013, and October 16, 2017, 170 patients over 18 years of age were enrolled prospectively and consecutively. These patients had been diagnosed with AC by orthopedic surgeons or rheumatologists and underwent shoulder radiographs and MRI.
Patients with MRI contraindications, prior shoulder surgery, shoulder osteoarthritis, calci c tendinosis, shoulder bursitis, and fractures on MRI were excluded. One patient withdrew from the study; four were excluded because of missing clinical data, 33 because of rotator cuff pathology (at least one fullthickness tendon tear). Thus, the nal study population consisted of 132 patients with a mean age 54.1±9.3  years. There were 55 men and 77 women (0.63 M/F sex ratio). Two patients had bilateral AC, yielding 134 shoulder MRI studies.

SHOULDER FUNCTION ASSESSMENT
A modi ed CMS was applied to all patients by a senior radiologist just prior to the MRI examination 19 .
Two subjective variables for a maximum score of 35 were evaluated: daily living pain (varying from 0severe pain to 15 points -no pain) and daily living activity limitation (varying from 0 -maximal limitation to 20 points -no limitation). The patients answered a questionnaire assessing the degree of pain (no pain, slight, moderate or severe pain), activity level (pain during work, sports, and recreation, sleep) and arm range of motion (ROM) (arm elevation up to the waist, xiphoid process, neck, top of the head, above the head). The examiner received prior training in performing the CMS. ROM was also quantitatively assessed with a goniometer, in external and internal rotation, forward and lateral elevation, and scored in each position by the examiner (varying 0°-30° = 0 to 151°-180°= 10 points for each movement). Thus, the ROM scored varied from 0 -minimal mobility to 40 -maximal mobility). The nal CMS, therefore, ranged from 0, (highly impaired shoulder) to 75 points (normal shoulder) (supplementary material 1). Shoulder strength, which was part of the original CMS, was not evaluated in this study, because there was no reliable measurement device available.
Pain duration was graded from 1-5 as follows: less than 6 weeks; between 6 weeks and 3 months; between 3 and 6 months; between 6 months and 1 year and over 1 year. The presence of diurnal pain, nocturnal pain, and nocturnal pain predominance were also evaluated.

CLINICAL FOLLOW-UP
A clinical follow-up was available in 49 patients with a mean age of 54±8.8 (37-74) years treated by physical therapy. There were 17 men and 32 women (0.53 M/F sex ratio). Based on medical record data (pain, activities, and ROM), the symptoms at 9 to 13 months after the MRI examination were classi ed as improved, stable, or worsened. None of these patients had been treated by intra-articular corticosteroid injection.

MRI EXAMINATION
MRI examinations were performed with either a 1.5T (105 patients) or a 3.0T scanner (27 patients

IMAGE ANALYSIS
The images were retrospectively reviewed by two musculoskeletal radiologists with three (FZ) and seven years (PP) of clinical experience with MRI blinded to clinical and demographic data using a PACS station (Synapse®, v4.1.600, Fuji lm, Montigny, France). A third radiologist (P.A.G.T.) with 11 years of clinical experience with MRI performed a training session with the two readers with 20 MRI studies of patients with AC, not included in the study population prior to the readouts.
The signal intensity of the IGHL on oblique coronal T2-weighted fat-saturated images was graded from 1-4 as follows ( Figure 1): normal homogenous low signal intensity; partial or foci of signal hyperintensity; global signal hyperintensity; linear hyperintensity of the peri-articular soft tissues.
The patients with IGHL scores of 1 and 2 were considered to have a low IGHL signal intensity, and those with grades 3 and 4 were considered to have high IGHL signal intensity. The thickness of the IGHL was measured at the glenoidal and humeral insertions on oblique coronal T2-weighted fat-saturated images 9 and classi ed as <4 mm, between 4 and <6 mm and ≥6 mm (Figure 2) 27 . The thickest portion of the coracohumeral ligament (CHL) was measured on the sagittal T2-weighted fat-saturated images 9 ( Figure  3).

STATISTICAL ANALYSIS
The R Development Core Team software (version 3.0.12013, R Foundation for Statistical Computing, Vienna, Austria) was used to perform statistical analysis. Statistical signi cance was de ned as P<0.05. Quantitative data are presented as mean ± standard deviation (range).
Linear regression analysis with the Pearson test was used to evaluate the correlation between the signs of AC studied on MRI and pain, mobility, activity scores, and pain duration. The association between MRI ndings, global modi ed CMS score, diurnal pain, night pain, and predominance of night pain was assessed with the Fisher exact test. The association between MRI ndings and clinical follow-up was assessed with the Wilcoxon test. For each MRI measurement, intraclass correlation coe cients (ICC) were calculated to assess interobserver variability. ICC values below 0.5 were considered poor, between 0.5 and 0.75 moderate, between 0.75 and 0.90 good and above 0.9 excellent 28 . Table 2 shows demographic characteristics, global modi ed CMS, and its items in the study population. The mean CMS score was 31.3±14.2 (2-69) points, and the mean pain duration grade was 3.5±1.1 (1-5) (table 3). Night pain was frequent, and predominant in about half of the concerned patients. Table 3 shows the pain duration grade in each grade of IGHL signal intensity. IGHL signal intensity was low in 70 shoulders (52.2%) and high in 64 (47.8%) for reader 1. These gures were 72 (53.7%) and 62 (46.3%), respectively, for reader 2. Table 4 shows the MRI ndings in the shoulders evaluated.

Results
ICC was excellent in grading IGHL signal as low or high (0.96), and moderate when taking in account all the four grades (0.67). ICC values were moderate for IGHL thickness (glenoidal insertion: 0.72, humeral insertion: 0.61) and poor for CHL thickness (0.09). Mobility scores were signi cantly different in patients with high IGHL signal intensity compared to those with low intensity for both readers (P = 0.04 and 0.02 for readers 1 and 2). The mean mobility scores between shoulders with low and high IGHL signal intensity grades were 19.6±10.1 (2-40) points versus 15.7±8 (0-38) points and 19.4±10 (0-40) points versus 15.8±8.2 (0-38) points respectively for readers 1 and 2. The variation of mobility scores with respect to IGHL signal intensity grade is shown in Figure 4.
There was a signi cant difference in pain duration for both readers (P=0.03 and 0.04 for readers 1 and 2) between patients with low and high IGHL signal intensity. The pain duration grades in patients with low and high IGHL signal intensity were 3.8±1.2 (1-5) versus 3.2±0.9 (1-5) and 3.8±1.2 (1-5) versus 3.2±0.9 (1-5) respectively for readers 1 and 2. As the IGHL signal intensity grade increased, there was also a decrease in mean pain duration grade for both readers (Table 3) ( Figure 5). For both readers, the highest frequency of high IGHL signal intensity was found in patients with pain lasting for 3-6 months. The highest frequency of low signal IGHLs were found in patients with more than one year of pain. The presence of high IGHL signal intensity was signi cantly associated with nocturnal pain predominance for both readers (P=0.003 and 0.003).
CHL measurements are shown in table 5. This ligament could not be measured con dently in 5 patients for reader 1 and 20 patients for reader 2. There was no association between CHL thickness and clinical impairment. For both readers, there was no association between CMS modi ed global score, pain intensity grade, diurnal pain, and MRI ndings.
Concerning clinical outcomes, 31 patients showed improvement, 11 patients stability and 7 worsening.

Discussion
Our study showed a signi cant correlation between high IGHL signal intensity and the pain duration in patients with AC, with a clear high signal predominance in the patients presenting pain from 3-6 months. Patients with the highest IGHL signal intensity were also shown to have the lowest score of pain duration.
Linear hyperintensity of the peri-articular soft tissues might be responsible for the moderate inter-observer agreement of IGHL signal intensity grading, but when IGHL signal was evaluated as low or high, reproducibility was excellent. High IGHL signal was also associated with night pain. Thus, capsular and pericapsular ligament signal in patients with AC is indicative of an early in ammatory disease stage, associated with in ammatory type pain and limited ROM. Additionally, our results showed that patients with a thick IGHL (4 mm or higher, regardless of measurement location) were likely to have a favorable outcome at follow-up (approximately one year). Conversely, thin IGHL (3 mm or lower) was associated with clinical worsening. Those results suggest that a thickened IGHL is related to an intense capsular in ammatory reaction, which seems to be associated with impaired shoulder function and a favorable prognosis. Hence, IGHL thickness may be seen as a prognostic biomarker in patients with AC, and further studies are necessary to assess the implication of this nding in patient management.
Our results about IGHL signal intensity are in agreement with Sofka et al. 20 , who stated that capsular high signal intensity in the axillary pouch was most closely associated with stage 2 disease, corresponding to pain from three to nine months. However, our ndings suggest that the high signal intensity is higher early in stage 2 disease. Capsular high signal intensity has been associated with stage 1 disease (no limitation of the ROM), however a signi cant decrease in mobility scores was seen in patients with high IGHL signal intensity, which has not been previously reported 24 .
Glenohumeral corticosteroid injections have been shown to be effective for short term pain relief and ROM improvement in the short and long term 29 . A recent study showed that signs of AC, including capsular edema and rotator interval signal abnormalities, were independent predictors of a better outcome for pain relief after glenohumeral corticosteroid injections 30 . These ndings indicate that high IGHL signal intensity re ects in ammatory joint reaction in early disease stages, suggesting that, for such patients, treatment with glenohumeral corticosteroid injection is more suitable than physiotherapy or that both treatments should be combined 31 .
No MRI nding was correlated with CMS, in agreement with Park et al. 32 , who showed a correlation of CMS with total arthrography score but not with MRI ndings. Unlike Anh et al. (25), we did not nd any correlation between MRI ndings and pain, but we did not rate IGHL enhancement. However, prior reports indicated that contrast enhancement does not improve the performance of IGHL T2 signal assessment for the diagnosis of AC 6 . Although, gadolinium injection has recently been shown to be helpful in di cult cases, contrast injection is not currently recommended for the evaluation of patients with AC 33,34 . IGHL thickness >4 mm has been described as a reliable sign of AC (27), but in our study, less than 30% of patients ful lled this criterion.
This study has limitations. Most importantly, AC diagnosis was con rmed neither by arthroscopy nor histologically. However, clinical ndings remain the basis for the diagnosis of AC, and the diagnostic performance of MRI diagnostic criteria has been previously evaluated [6][7][8][10][11][12]17,35,36 . IGHL and CHL thickness could vary depending on the coronal slice selected for measurement. As the estimation of pain duration provided by patients may be imprecise, a pain score system with time intervals was used to limit this potential bias. Since the correlation between MRI ndings and the range of motion each direction is still debated 24,32,35 , only global motion scores were considered. There was no control group, and no systematic clinical or MRI follow-up of the patients included. The possibility of a selection bias should be considered as our institution is a tertiary referral center, and patients with severe AC might have been over-represented. However, the study population one of the largest reported so far, with various disease stages and clinical impairment levels, minimizing this issue.
In conclusion, signal intensity at the IGHL was inversely related to pain duration and ROM. Patients with IGHL high signal were more frequently in the early phases of AC ( rst 3-6 months), presented nocturnal pain, and were likely to show clinical improvement on follow-up. Although IGHL thicknesses presented a moderate interobserver variability and an unclear relation with shoulder function impairment at the time of the MRI examination, there was a signi cant relation with patient prognosis. IGHLs >4mm were associated with a favorable clinical outcome, whereas ligaments measuring <3mm were associated with clinical worsening. These ndings should be considered in the MRI evaluation of patients with AC, and although further studies are still necessary, they could have therapeutic implications.  Diurnal pain, night pain and predominance of night pain are given for the 134 shoulders. Table 4. Pain duration grade according to inferior glenohumeral ligament signal intensity grade.

IGHL: inferior glenohumeral ligament
Results of pain duration score are presented on mean ± standard deviation. Range of all sub-groups of pain duration grade was 1-5.  Mean mobility score is shown according to inferior glenohumeral ligament intensity grade for reader 1 and reader 2, accompanied by a linear tendency (dotted line) curve for each reader. Note for reader 2 the v-shaped tendency, not found for reader 1. IGHL: inferior glenohumeral ligament.